Duke AI Health’s Nicoleta Economou talks guidelines & guardrails for responsible health AI development in AIMed “Champions” interview
Nicoleta Economou-Zavlanos, PhD, the director of Governance and Evaluation of health AI systems at Duke AI Health, was recently interviewed by AIMed’s Gemma Lovegrove for their AI Champions Interview Series, which highlights key thought leaders in the AI space. During the interview, Dr. Economou underscored the importance of incorporating fairness, transparency, and inclusivity throughout the entire process of health AI development, implementation, and monitoring.
The Connected Health Initiative (CHI) is hosting an in-person conference titled ‘Artificial Intelligence and the Future of Digital Healthcare at the Crossroads’ on September 26, 2023, at the National Press Club in Washington, D.C., from 12:30 PM to 5:35 PM EDT. The event will delve into the profound impact of AI systems on healthcare, offering potential for improved outcomes, cost savings, and a shift towards proactive disease prevention. Duke AI Health Director Michael Pencina, PhD, ABCDS Director Nicoleta Economou-Zavlanos, PhD, and AI Health Equity Scholar Michael Cary, PhD, RN, will be presenting the Algorithm-Based Clinical Decision Support (ABCDS) Oversight framework at the conference, touching upon the program’s design, implementation and strategies for bias mitigation and ensuring health equity. The CHI conference aims to foster a vital public dialogue on the state of health AI, proactive approaches by leading organizations to address AI efficacy, and the government’s role in managing AI’s risks and opportunities in healthcare.
Michael Pencina, PhD, vice dean for data science, professor of biostatistics and bioinformatics at Duke University School of Medicine, and director of Duke AI Health, has been named Duke Health’s first chief data scientist. Executive Vice President for Health Affairs and Dean Mary E. Klotman, MD, and Duke University Health System Chief Executive Officer Craig Albanese, MD, MBA, announced Pencina’s appointment. “In the current era of rapid expansion of AI and data science, we created this new role in recognition of the need for a well-articulated strategy for Duke Health that spans and connects both our academic and our clinical missions,” Klotman and Albanese said in their announcement.
Large language models (LLM) are powering amazing recent innovations in generative AI such as ChatGPT. Although their capabilities may seem like magic, behind these technologies are concepts that anyone can understand.
Please join us on Tuesday, September 19 for a lunch and learn as Larry Carin provides a math-free, intuitive explanation of how LLMs work. Dr. Carin will introduces participants to the deep-learning technology that has revolutionized the capacity of machines to perform language translation, to answer questions posed for given text, and to generate (synthesize) text that is near human-generated quality.
Duke University and Duke Health have recently announced a monumental five-year strategic partnership with Microsoft to support artificial intelligence (AI) applications in medicine and lead transformation in healthcare delivery, champion health equity, and pioneer advanced research. “We are excited to partner with Microsoft and bring our organizations’ talent together to solve the most pressing healthcare challenges,” said Duke AI Health Director and Vice Dean for Data Science Michael Pencina, PhD. “We will combine medical expertise, data science methods, and technology solutions to improve patient care and community health and advance the foundations of trustworthy health AI.”
Duke University Health System has received a new capacity-building contract with the Patient-Centered Outcomes Research Institute (PCORI). This contract, as part of PCORI’s Health Systems Implementation Initiative, will be used to support preparation for implementation projects that will advance the adoption of evidence-based practice within healthcare delivery settings. The Duke team, led by Rick Shannon, MD, includes co-investigators Armando Bedoya, MD; Nrupen Bhavsar, PhD; Ben Goldstein, PhD; and Michelle Lyn, MBA, MHA; and is supported by Duke AI Health.
Duke AI Health and School of Nursing to convene first-ever Duke Symposium on Algorithmic Equity and Fairness in Health
Duke AI Health and the Duke University School of Nursing are proud to announce the inaugural Duke Symposium on Algorithmic Equity and Fairness in Health, scheduled to take place in spring 2024.
The symposium will be spearheaded by Dr. Michael Cary, a distinguished scholar in nursing and the Elizabeth C. Clipp Term Chair of Nursing at the School of Nursing. Dr. Cary also serves as the Health Equity Scholar for Duke AI and leads the algorithmic equity initiative within Duke AI Health.
“In healthcare, algorithmic bias can lead to disparities in diagnosis, treatment recommendations, and access to care. It can disproportionately affect marginalized and underrepresented groups, exacerbating existing health inequities. As we rely increasingly on clinical algorithms to make decisions that impact people’s lives, we must continue to raise awareness about algorithmic bias in healthcare and work towards building a more equitable healthcare system,” stated Dr. Cary.
This groundbreaking symposium aims to bring together esteemed faculty members and experts from various disciplines to address bias resulting from clinical algorithms. The goal is to develop innovative methods and interventions that promote equity in health and healthcare delivery, particularly for marginalized groups. The event will revolve around the theme “Mitigating Bias and Advancing Health Equity in Clinical Algorithms in Healthcare.”
Algorithmic bias carries significant real-world implications that pervade various domains, including employment, housing, and healthcare. While many emerging methods are being employed to comprehend and mitigate algorithmic bias, critical gaps persist in the development and implementation of such vital approaches to advance health equity research and practice solutions.
Duke AI Health has prioritized algorithmic bias in health as a central focus of its mission to foster ethical and equitable data science. “I am thrilled to support Dr. Cary’s leadership in this essential domain and eagerly anticipate the expertise this event will bring together,” remarked Dr. Michael Pencina, Vice Dean for Data Science and Director of AI Health.
The Duke School of Nursing is deeply committed to mitigating the adverse social determinants of health and eradicating health inequities. ” I commend the efforts of Dr. Cary and the team at Duke AI Health in organizing the first Duke Symposium on Algorithmic Equity and Fairness in Health,” said Dr. Vincent Guilamo-Ramos, Dean and Bessie Baker Distinguished Professor in the Duke University School of Nursing. “This symposium will provide a valuable platform for experts to come together, share knowledge, and develop innovative solutions to advance health equity research and practice.” He went on to say, “I particularly encourage nurses to actively engage in these discussions and contribute to the ongoing efforts to create fair and unbiased algorithms in clinical settings and throughout the community where healthcare is delivered. Together, we can make a meaningful difference in promoting equity and fairness in healthcare.”
Save the date! The symposium is scheduled to take place in person at the Duke School of Nursing from March 13-14, 2024. Additional details and registration information will be announced in fall 2023. To stay informed about the event, we encourage individuals to sign up for the Duke AI Health mailing list and the Duke School of Nursing mailing list.
Sarah Riddle; Manager, External Communications, Duke University School of Nursing
Phone: (919) 613-9778
About Duke University School of Nursing
A diverse community of scholars and clinicians, Duke University School of Nursing is advancing health equity and social justice by preparing nurse leaders and innovators with a commitment to improving health outcomes through transformative excellence in education, clinical practice, and nursing science. Ranked as one of the leading nursing schools in the country, Duke School of Nursing focuses on improving the health of communities locally and globally by educating the nursing leaders of tomorrow and taking tangible steps to end health inequity, like the creation of www.DUSONtrailblazer.com, a set of conceptual and applied web resources for harmful social determinants of health mitigation.
About Duke AI Health
Duke AI Health is a pioneering initiative at Duke University that focuses on the ethical and equitable application of data science in healthcare. The mission of Duke AI Health is to drive innovation, research, and collaboration to advance health equity, improve patient outcomes, and transform healthcare delivery.
The Duke AI Health Data Science Fellowship Program is a 2-year training program in data science with direct application for healthcare. Designed for early-career data scientists with strong backgrounds in quantitative disciplines, the program is part of a multidisciplinary, campus-spanning initiative that applies machine learning and quantitative sciences to rich sources of healthcare and administrative data, using the insights gained to improve healthcare delivery, quality of care, and the health of individuals and communities.
Under the leadership of program director Lisa Wruck, PhD, and associate director Silvana Lawvere, PhD, the Data Science Fellowship program enrolled the first fellows in February 2020, and 12 fellows have participated in the program to date.
“Dr. Wruck and Dr. Lawvere have been integral to the success of this program,” said Michael Pencina, PhD, vice dean for data science and director of Duke AI Health. “Their expertise and commitment to the trainees has created a rigorous and supportive environment for them to learn and thrive, and I’m grateful to them for creating the success of the program.”
As Dr. Wruck and Dr. Lawvere step away from the program, the leadership will transition to Matt Engelhard, PhD as the faculty director and Andrew Olson, MPP as the senior operations leader. Dr. Engelhard is an Assistant Professor of Biostatistics and Bioinformatics and Mr. Olson is AI Health’s Associate Director, Policy Strategy and Solutions for Health Data Science.
A new video highlights the experience of the Duke Health Data Science Poster Showcase, held April 24, 2023 at the Mary Duke Biddle Trent Semans Center for Medical Education.
Created by Duke videographer Michael Blair with support from the Duke Center for Computational Thinking, the video features interviews by Matt Engelhard, PhD, an Assistant Professor of Biostatistics and Bioinformatics and faculty director of the AI Health Data Science Fellowship Program; and Hanxue Gu, a PhD student in the Electrical and Computer Engineering department and a member of the Mazurowski Lab who won the award for Best Computational Thinking Poster.
The Spring 2023 semester was comprised of 15 AI Health seminars that attracted 866 attendances, including people who attended multiple sessions.
The AI Health seminar series builds upon the success of our previous “Plus Data Science” learning experiences from 2017 – 2021 under the leadership of Larry Carin, which convened 59 in-person learning experiences and 66 virtual learning experiences.
Since its launch in 2018, +DS has held a cumulative 140 learning experience sessions (both in-person and virtual).
The Spring 2023 semester featured 32 data experts from across multiple areas of interest.
Duke Departments and Schools Represented:
- Biostatistics and Bioinformatics
- Duke Health Technology Solutions
- Biomedical Imaging
- Family Medicine and Community Health
- School of Nursing
- Civil and Environmental Engineering
- Electrical and Computer Engineering
- Computer Science
- Biomedical Engineering
- Population Health Sciences
- Office of Evaluation and Applied Research Partnership
We are looking forward to the Fall 2023 learning experiences, which will begin in September. A list of upcoming events can be found at https://aihealth.duke.edu/events/
Duke Researchers Develop Prediction Model to Identify Children With Complex Health Needs At Risk for Hospitalization
An important study led by Duke’s David Ming, MD, and AI Health’s Benjamin Goldstein, PhD, and Nicoleta Economou, PhD, on the use of predictive modeling to identify children with complex health needs who are at high risk for hospitalization, was recently published in Hospital Pediatrics, the official journal of the American, Academy of Pediatrics. The study analyzed data from electronic health records and found that certain demographic, clinical, and health service use factors were associated with a higher risk of future hospitalization. The authors, including Duke’s Richard Chung, MD, and Ursula Rogers, BS, suggest that the use of predictive modeling can help identify children with complex health needs who may benefit from targeted interventions to prevent hospitalizations and improve outcomes. The study is accompanied by a commentary by University of Wisconsin Neil Munjal, MD, MS, titled ‘Machine Learning: Predicting Future Clinical Deterioration in Hospitalized Pediatric Patients,’ which describes the Duke researchers’ machine learning approach as “thought-provoking.”
A poster showcase held on Monday, April 24, 2023 at the Mary Duke Biddle Trent Semans Center for Medical Education featured 28 posters in health data science. This cross-disciplinary event was hosted by multiple organizations, including Duke AI Health, the Laboratory for Transformative Administration, and the Center for Computational Thinking. Poster topics were centered around health data science and covered a wide range of topics including statistics, informatics, machine learning, data engineering, implementation, process engineering, technology development, and applications. The posters were submitted by people at all stages of their careers, including students, trainees, staff, and faculty. Information tables also shared programs and resources relevant to health data science at Duke.
The Health Data Science poster showcase will be held in person on Monday, April 24 from 12:00-2:00 PM. We’re excited that this cross-disciplinary event will be hosted by multiple organizations, including Duke AI Health, the Laboratory for Transformative Administration, and the Center for Computational Thinking.
The poster display will take place in the Mary Duke Biddle Trent Semans Center for Medical Education (Trent Semans) on the 6th floor and we’ll serve light refreshments.
More than 25 posters will be presented, including Duke participants from: AI Health Fellowship Program; Biomedical Engineering; Clinical and Translational Science Institute (CTSI); Computer Science; Department of Biostatistics and Bioinformatics; Department of Internal Medicine; Department of Neurosurgery; Department of Surgery; Division of Geriatrics, Department of Medicine; Division of Hematology, Department of Medicine; Duke Clinical Research Institute (DCRI); Electrical and Computer Engineering (ECE); Duke Health Technology Solutions (DHTS); Laboratory for Transformative Administrative (LTA); Master of Management in Clinical Informatics (MMCi); OB-GYN; and Trinity College of Arts & Sciences.
Information tables will include programs from across Duke: The + Programs for Students; Duke AI Health; Biostatistics, Epidemiology, and Research Design (BERD); Center for Computational Thinking; Duke Data Analytics Community; and the Master of Management in Clinical Informatics.
Poster awards will include Best Computational Thinking Poster and the Good DEEDS Award (Ethical and Equitable Data Science).
Please join us! All are welcome, and light refreshments will be served.
The Coalition for Health AI (CHAI) released its highly anticipated “Blueprint for Trustworthy AI Implementation Guidance and Assurance for Healthcare” (Blueprint). The Blueprint addresses the quickly evolving landscape of health AI tools by outlining specific recommendations to increase trustworthiness within the healthcare community, ensure high-quality care, and meet healthcare needs. The 24-page guide reflects a unified effort among subject matter experts from leading academic medical centers and the healthcare, technology, and other industry sectors, who collaborated under the observation of several federal agencies over the past year.
Duke University Health System has been selected by the Patient-Centered Outcomes Research Institute (PCORI), an independent, nonprofit research organization, to participate in a new effort to close the gap between high-quality medical research and implementation of that evidence to improve patient outcomes.
The PCORI project will support ongoing efforts at Duke centered on work begun through the Duke Quality System. Led by Richard P. Shannon, M.D, Duke Health senior vice president and chief quality officer, the Duke Quality System aims to provide “perfect patient care,” a concept that not only includes providing timely, evidence-based patient care, but also ensuring that the care is done right the first time, without defects, waste, or inequity.
Shannon, who also serves as chief medical officer for Duke Health, most recently has led the development of the Duke Collaborative to Advance Health Equity (CACHE), a community-driven program that extends the quality system model by harnessing data science to find and eliminate racial disparities in health care.
The Health Data Science poster showcase will be held on Monday, April 24 from 12:00-2:00 PM in-person in the Mary Duke Biddle Trent Semans Center for Medical Education (Trent Semans). We’re excited that this cross-disciplinary event will be hosted by multiple organizations, including Duke AI Health, the Laboratory for Transformative Administration, and the Center for Computational Thinking.
We invite any member of the Duke community to propose a poster entry for participation in this event, including students, trainees, staff, and faculty. This experience is intended to be especially valuable to individuals seeking to gain experience in presenting their work in front of a scientific audience, and the poster itself can become a valuable part of an academic portfolio.
Submit your poster topic at: https://duke.qualtrics.com/jfe/form/SV_1HuvnGKOY4YMa9w
The preferred deadline for poster topics to be submitted is Monday, March 13, 2023 by 11:59 PM (Eastern time).
Update on March 10: We’ve heard from several people that their research is ongoing, and we’ve decided to accept poster topics on an rolling basis, to allow everyone the full opportunity to participate.
Poster topics must be centered around health data science, but can cover a wide range of potential topics, such as statistics, informatics, machine learning, data engineering, implementation, process engineering, technology development, or applications. We especially encourage submissions describing experiences with Duke data sources. Student posters describing class projects (at both the undergraduate and graduate levels) are also encouraged.
After you submit your topic, you’ll then receive a poster template with the correct dimensions.
You’ll need to submit your finalized poster by Friday, April 14 in order to have it printed. If your poster is accepted, the event organizers will print it for you and you will have no cost to participate. The showcase will include poster judging, with recognitions including best poster.
The fall 2022 semester was comprised of 9 AI Health seminars that attracted 463 attendances, including people who attended multiple sessions. Across all of 2022, the AI Health seminar series has hosted 22 virtual seminars with 1,820 cumulative attendances. The AI Health seminar series builds upon the success of our previous “Plus Data Science” learning experiences from 2017 – 2021 under the leadership of Larry Carin, which convened 59 in-person learning experiences and 66 virtual learning experiences. Since its launch in 2018, +DS has held a cumulative 125 learning experience sessions (both in-person and virtual). – Metrics by Tiffany Torres
Current medical standards for accessing stroke risk perform worse for Black Americans than they do for white Americans, potentially creating a self-perpetuating driver of health inequities. A study, led by Duke Health researchers and appearing online Jan. 24 in the Journal of the American Medical Association, evaluated various existing algorithms and two methods of artificial intelligence assessment that are aimed at predicting a person’s risk of stroke within the next 10 years. The study found that all algorithms were worse at stratifying the risk for people who are Black than people who are white, regardless of the person’s gender. The implications are at the individual and population levels: people at high risk of stroke might not receive treatment, and those at low or no risk are unnecessarily treated.
Last fall AI Health held an in-person workshop designed to give hands-on experience in working with medical digital pathology images using machine learning. See highlights from the afternoon in a video created by our partners in the Center for Computational Thinking. The concept of “do machine learning in just one afternoon!” was very successful, and we appreciate the participation from all those who attended. We are currently working to design more such studios, and please join our mailing list if you’d like to be notified for upcoming events.
AI Health Seminar: ABCDS Oversight – A framework for the governance and evaluation of algorithms to be deployed at Duke Health
Save the date: February 14, 2023, 12:00 PM EST: Duke AI Health’s Nicoleta Economou, PhD, joins Duke DHTS’s Armando D. Bedoya MD MMCi, to present: ‘Algorithm-Based Clinical Decision Support (ABCDS) Oversight: A framework for the governance and evaluation of algorithms to be deployed at Duke Health.’ During the webinar, which is open to members internal and external to Duke, Drs. Economou and Bedoya will discuss highlights from their recent paper published in the Journal of the American Medical Informatics Association (JAMIA).
Recent years have seen growing interest in the use of artificial intelligence tools for healthcare applications, including diagnosis, risk prediction, clinical decision support, and resource management. Capable of finding hidden patterns within the enormous amounts of data that reside in patient electronic health records (EHRs) and administrative databases, these algorithmic tools are diffusing across the world of patient care. Often, health AI applications are accompanied by assurances of their potential for making medical practice better, safer, and fairer. The reality, however, has turned out to be more complex.
Duke AI Health’s HDS research and education hub held a successful Poster Showcase on December 6, 2022, featuring the work of 16 students and fellows. Hosted by Ricardo Henao, PhD, and Shelley Rusincovitch, MMCi, the presenters included members of the HDS fall 2022 student cohort, fellows in the AI Health Data Science Fellowship program, as well as members of AI Health’s Spark Imaging Initiative and Duke Biostatistics & Bioinformatics’s BCTIP program.
Congratulations to AI Health Faculty Council member Ben Goldstein, PhD, and Duke Children’s Health & Discovery Initiative Director Jillian Hurst, PhD, for their success in leading the Electronic Health Record (EHR) Study Design Workshop from December 5-9, 2022. The course was offered as a virtual 5-day class providing foundational lectures and hands-on studios on the fundamentals of working with, and designing EHR-based studies. The inaugural workshop generated a great deal of enthusiasm and every seat in the course was filled within 6 weeks of course announcement.
Join Duke AI Health Director Michael Pencina, PhD, as he takes part in discussions with expert panelists convened from government, industry and academia to discuss recent advances in health AI, including structural biological modeling, computer vision algorithms, and ethical frameworks for employing AI in healthcare. This virtual event, “Modeling Equitable AI in Digital Health,” is hosted by MITRE and will take place starting at 4:00 PM EST on Thursday, December 8, 2022.
DCRI Science and Digital Officer Eric Perakslis, PhD, shares a deeply personal perspective on a recent federal mandate that expands patients’ access to data stored in their EHRs – but also carries its own potential for risks. In his essay, Dr. Perakslis combines the patient and tech expert viewpoints as he surveys the “lumpy, bumpy, imperfect progress” toward better data transparency while undergoing cancer diagnosis and treatment.
Duke AI Health is pleased to announce the Duke Electronic Health Records Study Design Workshop (EHR-SDW) 2022. The workshop will be offered in December as a virtual five-day class that provides foundational lectures and hands-on studios on the fundamentals of working with and designing EHR based studies. The EHR-SDW is targeted toward individuals interested in learning about how to work with and conduct studies using electronic health records (EHR) data. EHR data are a widely available form of real-world data that have become standard in studies ranging from clinical trials, comparative effectiveness, risk prediction, and population health. This workshop is offered through Duke AI Health’s Health Data Science (HDS) program and builds on the success of our highly successful Machine Learning Schools, with 11 events held since 2017.
The AI Health Data Science Fellowship Program is a two-year training program focused on data science with healthcare applications, designed for early-career data scientists with strong backgrounds in quantitative disciplines. Launched in fall of 2019, the program currently has 5 fellows, 2 staff data scientists, and 5 alumni. The program recently came together in-person for lunch for the first time since the pandemic. They gathered to welcome 2 new members: new fellow Angel Huang and new Data Scientist, John Rollman.
This in-person workshop presented by Ricardo Henao, PhD; Associate Professor, Department of Biostatistics and Bioinformatics; Chief AI Scientist, Duke AI Health, Akhil Ambekar, MS; Fellow, AI Health Data Science Fellowship Program, with Shelley Rusincovitch, MMCi; Managing Director, Duke AI Health, will give you hands-on experience in working with medical digital pathology images using machine learning. Our use case will be in whole slide images of lymph node sections. We will use the CAMELYON16 dataset (https://camelyon16.grand-challenge.org/), which consists of 400 hematoxylin and eosin-stained whole-slide images. During the workshop, you will learn how to retrieve, manage, and process these images, then apply a machine learning model based on a neural network architecture to classify image regions as normal or malignant. The techniques you learn will also be broadly applicable to other types of medical imaging.
Duke AI Health congratulates Chief AI Health Scientist Ricardo Henao, PhD, on his promotion to the rank of Associate Professor in the Department of Biostatistics and Bioinformatics in the Duke University School of Medicine. Dr. Henao is a major presence in health data science at Duke, where his leadership and expertise in machine learning methods and implementation have made him a sought-after collaborator and instructor. “Dr. Henao is a major asset to Duke AI Health and to the larger Duke community,” said Michael Pencina, PhD, director of Duke AI Health and vice dean for data science at the School of Medicine. “We feel fortunate to be able to benefit from such a rare combination of talent and knowledge spanning research, application, and teaching.”
Duke AI Health Director and Vice Dean for Data Science Michael J. Pencina, PhD, has achieved a major academic milestone: according to Google Scholar’s analytics, he has recently passed the 100,000 mark for academic citations of his work. Pencina, who in addition to his leadership role in Duke’s efforts to develop, evaluate, and implement ethical and equitable data science, has also worked extensively on the development and evaluation of risk prediction models and clinical trial designs.
As a member of the Coalition for Health AI, Duke AI Health is working to develop a consensus-driven framework to drive high-quality health care through the adoption of credible, fair, and transparent health AI systems. The coalition is convening a series of virtual workgroup sessions to define core principles and has published a white paper from its first meeting: “Bias, Equity, and Fairness.” Please review the paper and submit your feedback by Sept. 15: https://bit.ly/3wbAXQx. With the help of your ideas, the Coalition for Health AI can advance towards establishing clear and appropriate guidelines and guardrails for the fair, ethical, and useful application of AI and machine learning in health care settings.
Much-Touted Genomic Test Score Shows Minimal Utility in Study Led by AI Health Director Michael Pencina
New research led by Duke AI Health Director Michael Pencina, PhD, published recently in the journal Circulation, looked at the value of using a genomic test to predict the future risk of heart disease. Pencina and colleagues found that the genomic test, referred to as the polygenic risk score (PRS), only marginally added to the predictive information obtained through the assessment of traditional risk factors, concluding that the PRS “had minimal clinical utility”.
We invite Duke students to apply for the Health Data Science (HDS) fall research program. This competitive program, based in Duke AI Health, is designed to allow students who have previous experience in data science to continue their engagement with substantive applied projects. The HDS Research Program offers Duke students, both undergraduate and graduate, the opportunity to be a part of research teams applying advanced machine learning (deep learning) to important areas of medicine. Participating students will be mentored by leading Duke faculty involved in data science research, often with guidance by practicing clinicians. The fall will culminate in a showcase session where student teams will present their results.
Ben Goldstein and Jillian Hurst share their experiences developing the Clinical Research with Electronic Health Records (CR-EHR) to bring together investigators with different backgrounds to learn how to collaborate on the design and execution of EHR-based studies.
A group of Duke Health researchers recently shared their insights on approaches to managing the complex issues that are emerging as “algorithmic medicine” increasingly becomes part of clinical care at hospitals and health systems. The authors, who comprise faculty and staff from Duke AI Health, Duke Health Technology Solutions, the Duke Institute for Health Innovation, and other physicians and researchers from Duke University and Duke University Health System, published an account of their approach to evaluating and monitoring the use of algorithmic predictive models at Duke Health hospitals and clinics. The article, titled “A framework for the oversight and local deployment of safe and high-quality prediction models,” was published on May 31 in the Journal of the American Medical Informatics Association (JAMIA). It showcases the processes and procedures by which an expert group at Duke Health known as Algorithm-Based Clinical Decision Support (ABCDS) Oversight reviews, approves, and manages predictive models intended for use in patient care settings.
Can AI safely automate medical decision-making tasks to improve patient outcomes? In this talk, the presenters will share the challenges in the development and translation of medical AI, and how they are being addressed through a blend of innovation in algorithm development, dataset curation, and implementation design. They will first talk about self-supervised learning methods for medical image classification that leverage large unlabeled datasets to reduce the number of manual annotations required for expert-level performance. Then, they will discuss open benchmarks that can help the community transparently measure advancements in generalizability of algorithms to new geographies, patient populations, and clinical settings. Third, they will share insights from studies that investigate how to optimize human-AI collaboration in the context of clinical workflows and deployment settings. Altogether, this talk will cover key ways in which we can realize the potential of medical AI to make healthcare more accurate, efficient and accessible for patients worldwide.
The Duke+Data Science program is pleased to announce the Duke Machine Learning Summer School 2022, offered in June as a live five-day class that provides lectures on the fundamentals of machine learning. The curriculum in the MLSS is targeted to individuals interested in learning about machine learning, with a focus on recent deep learning methodology. The MLSS will introduce the mathematics and statistics at the foundation of modern machine learning, and provide context for the methods that have formed the foundations of rapid growth in artificial intelligence (AI).
We invite Duke students to apply for the Health Data Science (HDS) summer research program. This competitive program, based in Duke AI Health, is designed to allow students who have previous experience in data science to continue their engagement with substantive applied projects. The Advanced Machine Learning Projects in Health Data Science offers Duke students, both undergraduate and graduate, the opportunity to be a part of research teams applying advanced machine learning (deep learning) to important areas of medicine. Participating students will be mentored by leading Duke faculty involved in data science research, often with guidance by practicing clinicians. The summer will culminate in a showcase session where student teams will present their results.
Duke AI Health welcomes Maciej Mazurowski, PhD, who will join its Faculty Council as Director of Radiology Imaging. At AI Health, Dr. Mazurowski will coordinate the AI Health Initiative for Medical Imaging. This new effort will engage experts in machine learning and clinical medicine from across Duke’s campus to foster and accelerate the development, validation, and clinical implementation of machine learning algorithms for medical imaging. “I’m excited to undertake this new challenge and I’m looking forward to working with experts and leadership across the entire campus to build on existing technical and clinical strengths in medical imaging AI at Duke,” Dr. Mazurowski said.
The mission of Duke AI Health is to enable the discovery, development, and implementation of artificial intelligence (AI) at Duke and beyond. A key component to achieving this goal is to foster high-impact, rigorous, and competitive proposals for scientific awards. The 2022 AI Health Proposal Studios will provide a structured opportunity for investigators to engage with Duke’s top data science expertise and thought leadership, and to receive review and feedback of the scientific components of their proposals. After seeing a strong response to the Proposal Studio concept and the following virtual learning experiences in 2021, AI Health plans to continue building on last year’s success with the overarching goal of fostering high-impact, rigorous, and competitive proposals for scientific awards.
In a recent post at the Tableau blog, the data visualization company praises the Duke Analytics Community (DAC) for the group’s commitment to “taking data democracy to (the) next level.” The post, which is available at the Tableau website, singled out the Duke Cancer Institute’s Claire Howell and Duke University’s Rebecca McDaniel for recognition based on their initiative in helping to create a “department-agnostic space” where users of analytics software across the School of Medicine and Health System could share ideas and improve data access.
Duke AI Health welcomes its first AI Health Equity Scholar, Michael P. Cary, PhD, RN, who is now beginning a yearlong scholarship supported by Duke AI Health and the Duke Clinical & Translational Science Institute. The AI Health Equity Scholars Program, which provides funding for Duke University faculty, staff, and postdoctoral scholars to actively collaborate with AI Health leadership, is focused on broadening Duke’s commitment to ethical and equitable data science and artificial intelligence (AI) in health applications.
Duke AI Health is pleased to launch the AI Health Data Studio Seminar series this spring. This multi-part educational offering is designed for campus-based researchers at Duke who are interested in working with medical data but are unsure where to begin. Hosted by Senior Informacist Ursula Rogers, Chief AI Health Scientist Ricardo Henao, PhD, and Associate Director of Informatics Shelley Rusincovitch, MMCi, the series will feature data experts from across the Duke enterprise.Campus-based researchers are especially invited to attend along with anyone interested from the Duke community, including faculty, staff, and students.
Ursula Rogers shares a deeply personal perspective on medical data. “No patient or family should ever have to worry about whether bad data might be steering medical care in the wrong direction, or whether good data is failing to reach a care provider who can act on it.”
Duke AI Health and the Duke Clinical & Translational Science Institute are pleased to announce a call for applications with the AI Health Equity Scholars Program. This program will support a minimum 1-year appointment for a faculty member, staff member, or postdoctoral scholar at Duke University. The AI Health Equity Scholars Program is a new initiative intended to broaden our commitment to ethical and equitable data science and artificial health (AI) applications, with direction from CTSI Director L. Ebony Boulware, MD, MHS, and AI Health Director Michael J. Pencina, PhD. The intention of this program is to broaden our expertise in considering and applying ethical and equitable principles for key initiatives within Duke AI Health. Applications must be submitted by Friday, December 10, 2021 by 10 PM (Eastern Time).
Given the rapid growth in and importance of harnessing health data as a tool, Mary Klotman, MD, Dean, Duke University School of Medicine, recently announced the key leadership appointment of Michael Pencina, PhD, Vice Dean for Data Science for the School of Medicine, as the Director of Duke AI Health effective October 13, 2021. Designed as a multidisciplinary initiative, AI Health intends to unlock the enormous opportunity to spur collaborations that will leverage knowledge and expertise from across campus.
Duke AI Health and the Duke Clinical & Translational Sciences Institute are pleased to announce a call for applications to the Spring 2022 Clinical Research with Electronic Health Records Data (CR-EHR) Course, with a November 19, 2021 application submission deadline. CR-EHR is an interdisciplinary course designed to engage both clinical and quantitative researchers in learning how to access and work with Duke EHR data. Data captured in the Duke EHR represent the broad spectrum of patient care delivered by Duke Health, which can be leveraged for a variety of research questions and study designs. Clinical trainees will develop a deeper understanding of the types of analytic studies that can be conducted with EHR data, while quantitative trainees will develop a deeper knowledge base for how to query and process EHR data.
Now more than ever, clinicians can access an incredible amount of data about their patients. Electronic health records (EHRs) offer a massive repository of information about each individual: notes of all kinds, laboratory results, imaging data, scanned forms, and saved images. Soon, we may even be able to add data from wearable devices such as personal fitness trackers into the mix. However, this breadth of information can be both a blessing and a curse. Clinicians can learn more about their patients from the medical chart than was previously possible—but only if they are able to rapidly and accurately sort through that information and find the most relevant points for a given clinical encounter.
This December, Duke AI Health Director and Professor of Biostatistics and Bioinformatics Michael Pencina will join a group of experts for a panel discussion hosted by the Duke Alumni Forever Learning Institute called “Artificial Intelligence: Capabilities, Liabilities, and Responsibilities.” The discussion, the final installment in a four-part series taking place this fall titled “Artificial Intelligence: Real Ethical Quandaries,” will focus on the expanding role of artificial intelligence in decision-making and the practical and ethical issues that can arise from the use of a technology whose inner workings are often opaque and whose operations can be affected by built-in biases. Panel participants will examine how these technologies are being used in arenas such as medicine and national security and discuss the potential impacts of these tools, both positive and negative, on people’s daily lives. The session will take place as an online Zoom webinar on Tuesday, December 7, 2021, from 3:00-4:00 PM Eastern time, and will be moderated by Duke Law Professor and Director of the Duke Initiative for Science and Society Nita Farahany.
Eric Perakslis, PhD, DCRI’s Chief Science & Digital Officer, will present at DEF CON 2021 in a talk called “Truth, Trust, and Biodefense.” Learn more about his presentation in his blog post for the DCRI below:
“On May 12, 2017, a ransomware cyberattack known as WannaCry was launched. Within a day, it was raging worldwide and had infected tens of thousands of computers and electronic devices belonging to the United Kingdom’s National Health Service, causing severe disruptions to hospital operations. Shortly after 15:00 UTC on May 13, the infection was halted when information security researcher and hacker Marcus Hutchins discovered and exploited a “kill switch” embedded in the malware’s code. In addition to greatly slowing WannaCry’s spread, this kill switch also prevented infected computers from being encrypted and their data locked. Marcus Hutchins’ story is notably complex, but there is no denying that his actions greatly decreased the global harm that likely would have otherwise occurred. The term hacker often brings to mind a faceless, hooded figure that is ubiquitously linked to crime. Given how pervasive this image is, it may surprise some to learn that there are many “good” hackers. This distinction is made especially clear in the viral TED Talk given by cybersecurity Keren Elazari titled “Hackers: the internet’s immune system.” In this talk, Elazari argues that hackers make the internet stronger by testing its defenses, which forces the internet to adapt, improve, and strengthen, not unlike the body’s adaptive immune system.
Duke AI Health Director Michael Pencina, PhD, who is a professor of biostatistics and bioinformatics at Duke and serves as the medical school’s vice dean for data science and information technology, was recently quoted in an article appearing in STAT News examining the use of commercially developed predictive algorithms in medicine. In an investigative report for STAT News, correspondent Casey Ross spoke with employees in multiple health systems across the country that use clinical algorithms created by Epic, one of the nation’s largest electronic health record vendors.
Duke+DataScience (+DS) is a Duke-wide educational initiative devoted to expanding knowledge of and facility with machine learning and other artificial intelligence tools across multiple academic fields, including the arts, humanities, and social sciences as well as medicine and quantitative sciences. With an extensive and growing curriculum that includes both online and in-person courses in neural networks, natural language processing, deep learning, and other machine learning applications, +DS offerings span learning needs ranging from novice to expert and are tailored to specific academic and professional applications.
Duke’s +Data Science (+DS) recently concluded its 2021 Machine Learning Virtual Summer School (MLvSS). This event, the ninth machine learning school held since 2017, sold out more than a month in advance and completely filled a 100-person waitlist. This high demand reflects both the substantial demand for instruction in methods driving the rapid growth in artificial intelligence, as well as a keen interest in tapping into high-quality instruction from Duke teachers with expertise in the mathematics and statistics that underlie modern machine learning methods.
Keeping up with the pace of research in health data science is challenging at the best of times, and the COVID-19 pandemic has not made things any easier. For this reason, Duke AI Health and the Duke +Data Science (+DS) program worked together this spring to launch the Proposal Studio Virtual Learning Experiences (vLE). The Proposal Studios sessions were designed to help investigators develop effective, successful proposals for research project involving health data science. From March through April of 2021, +DS held four successful proposal studios, assisting 13 investigators to develop scientific proposals. Open to anyone within the Duke community, the series attracted a total of 129 attendees and averaged 32 audience members per vLE.
Duke Rheumatologists Explore the Effects of a Rapid Transition to Telemedicine During the COVID-19 Pandemic
The COVID-19 pandemic has prompted a surge in demand for telehealth services, but many questions about how healthcare providers can adapt their practice to meet the challenges of telemedicine remain to be answered. Now, a group of rheumatologists at Duke University School of Medicine have used data drawn from the Duke University Health System’s EHRs (electronic health records) to investigate how a rapid transition to telemedicine affected their approach to patient care.
A group of neuroscientists and machine learning experts are developing new ways to analyze animal movement and behavior to gain insights into the inner workings of the nervous system. Combining expertise from the disciplines of neurobiology and artificial intelligence, a team of researchers from Duke University, Harvard, MIT, Rockefeller University, and Columbia University have developed a system that captures detailed, multiple-view video of animals in their natural environment, and then uses data from those video images to build a detailed model of how the animal moves. This allows scientists to use movement and behavior as a window into brain function.
Ursula Rogers, senior informaticist with Duke Forge and AI Health, recently presented a poster at the American Medical Informatics Association (AMIA) 2021 Virtual Informatics Summit. The poster, “Enabling Data Liquidity for Health Data Science: A Suite of APIs for EHR Data” discusses an ongoing partnership between the Duke Health Technology Solutions (DHTS) Analytic Center of Excellence and AI Health. 18 application programming interfaces (API) have been developed to provide efficient and secure programmatic access to electronic health record (EHR) data for machine learning.
Since it was declared a global pandemic in March 2020, COVID-19 upturned university and college campuses across the United States, causing major disruption to student life. As Duke’s campus went into a full lockdown following a steep uptick in COVID-19 infections in North Carolina last spring, Duke’s Harshavardhan (Harsha) Srijay, a 19-year-old second-year undergrad student majoring in math and data science, saw his plans for the 2020 summer crumble. As prior opportunities fell through the cracks, the Duke Plus Data Science (+DS) Advanced Projects summer program provided him a platform to not only be engaged and productive through a very difficult summer, but also come out of it with a successful project that he recently presented at the American Medical Informatics Association (AMIA) 2021 Virtual Informatics Summit(link is external).
In this one-hour virtual learning experience, 3 teams of Duke investigators will discuss their proposal concepts with data science experts. For April 5, proposal concepts will include genomic analysis related to sickle cell anemia, lifestyle intervention adherence, and transplant optimization. The proposal studio vLE concept is newly launching in spring 2021, with the goal of assisting Duke investigators with proposal development in health data science, and in sharing experiences with the broader Duke community. The series is co-hosted by Duke AI Health and the Duke+Data Science (+DS) program.
Now Accepting Proposals for Placement of a Pathology AI Health Fellow for Projects within the Department
AI Health is currently considering requests for placement of a Pathology AI Health Data Science Fellow. The AI Health Data Science Fellowship is a 2-year training program in data science with direct application for healthcare. The Pathology AI Health fellow will be funded jointly by AI Health and the Department of Pathology. Fellows will also receive support from AI Health and the Duke Department of Biostatistics and Bioinformatics, with overall program supervision provided by the Duke Clinical Research Institute’s Center for Predictive Medicine.
Now Accepting Proposals for Placement of a Microsoft–Duke AI Health Fellow for Projects within the School of Medicine
AI Health is currently considering requests for placement of a Microsoft-Duke AI Health Data Science Fellow for projects proposed by Departments/Divisions within the Duke University School of Medicine. The Microsoft–Duke AI Health Data Science Fellowship is a 2-year training program in data science with direct application for healthcare. Funded in part by a grant from the Microsoft Corporation, Microsoft-Duke AI Health Fellows will also receive support from AI Health, the clinical divisions to whose projects they are assigned, and the Duke Department of Biostatistics and Bioinformatics, with overall program supervision provided by the Duke Clinical Research Institute’s Center for Predictive Medicine.
The mission of Duke AI Health is to enable discovery, development, and implementation of artificial intelligence (AI) at Duke and beyond. A key component to achieving this goal is to foster high-impact, rigorous, and competitive proposals for scientific awards. The AI Health Proposal Studios will provide a structured opportunity for investigators to engage with Duke’s top data science expertise and thought leadership, and to receive review and feedback of the scientific components of their proposals. The deadline for submitting applications is 5:00 PM Eastern time on Monday, December 7, 2020.
The COVID-19 pandemic has produced a staggering array of challenges that clinicians, public health experts, and policy makers are struggling to meet. Data scientists and quantitative experts across the globe have gone into overdrive as they work to analyze a flood of information, seeking not only to better understand, track, and predict the disease, but also to help guide the response to it and ensure that timely, accurate, and trustworthy information is readily available for everyone from scientists and clinicians to communities and members of the public.
This urgent need to bridge the worlds of data science, clinical research, and public health was the driving force behind this summer’s COVID + Data Science Virtual Seminars. Sponsored by Duke Plus Data Science (+DS), the 8-week series in summer 2020 was devoted to exploring data science methods with direct applications to the COVID-19 pandemic.
The series of 12 lectures attracted more than 1,500 virtual attendances, with many participants joining across multiple weeks and topics. The series, which wrapped up in late August, provided the opportunity for audiences both within Duke and beyond to hear directly from experts on topics spanning data analysis and visualization, deep learning, statistical methods, natural language processing, molecular methodology, and more.
Read the full story at the Center for Computational Thinking website: https://computationalthinking.duke.edu/2020/11/03/plus-ds-covid-2020-summer-seminars/
A Machine Learning for Mobile Health workshop, part of the upcoming Neural Information Processing Systems Conference (NeurIPS 2020), is inviting contributions and extended abstracts from researchers and clinicians in the interdisciplinary machine learning and mobile health space, with the goal to better address the various challenges currently facing the widespread use of mobile health technologies in health and healthcare. Co-organized by Duke Statistical Science assistant professor Katherine Heller, PhD who is also a research scientist at Google AI, the workshop aims to facilitate collaboration between machine learning researchers, statisticians, mobile sensing researchers, human-computer interaction researchers, and clinicians from around the world.
Seven Duke +DS learning experiences will be held in September. These sessions offer the opportunity to dive deeper into topics and target diverse units at Duke: from those that desire a broad understanding of what is possible with data science, and those who wish to use data-science tools (software) without a need for deep understanding of underlying methodology, to those who desire a rigorous technical proficiency of the details and methodology of data science. Anyone in the Duke community is welcome to join, there is no fee to attend, and no prior experience is necessary.